An, P.E., Harris, C.J., Tribe, R. and Clarke, N.
Aspects of Neural Networks in Intelligent Collision Avoidance Systems for Prometheus.
Joint Framework for Information Technology
Full text not available from this repository.
This paper presents our work on adaptive driver modelling and obstacle classficiation applications, which will be incorporated into an intelligent collision avoidance system (ICAS) for road vehicles. The reliability of the ICAS is largely determined by the accuracy of these models. Multi-layered-Perceptron and Cerebellar-Model-Articulated-Controller neural networks were used in constructing the driver and obstacle classification models, and were evaluated using a car-following scenario for the driver model and a two-class obstacle (car or pedestrian) for the classification model. In the driver modelling application where the input dimension was low and training samples were rich, the CMAC network was found to achieve better accuracy than the MLP network. On the other hand, in the obstacle classification application where the input dimension was high and training samples were sparse, the MLP network was found to have fewer classification errors than the CMAC network. In both cases, the CMAC network converged significantly faster than the MLP network.
Actions (login required)